The era of AI is surging, and tool-based software is facing unprecedented challenges and changes. From what was once a small and beautiful tool to now struggling to survive in the AI torrent, they are either acquired or eliminated. In this “battle royale” of tool software, how to find a way to survive has become a problem that every tool software manufacturer must face.
Many small but beautiful tool-based software are embarrassing to develop in the end.
Evernote is one such note-taking app.
Evernote was founded in 2000, and the company’s founder, Stepan Pachikov, gave this product a classic positioning – Evernote is an extension of the human brain.
The mission of this product is to help users remember everything in life and accomplish anything in order to compensate for and overcome the natural limitations of human memory.
The product quickly became popular all over the world. Evernote entered the Chinese market in 2012 and operates independently in China under a well-known name, Evernote, to better localize its products.
In 2014, Evernote surpassed 100 million users, making it the world’s largest note-taking app.
But over the next decade, Evernote’s performance has been hard to describe.
They made a strategic mistake by overextending the product’s non-core features. Around 2014, Evernote launched edge products such as Evernote Food (food record), Hello (network management), and Market (peripheral e-commerce) in an attempt to maintain the image of an innovator through media exposure.
The maintenance costs of these non-core features take up more than 40% of R&D resources, while Evernote does nothing about the core features.
From 2014 to 2018, the development of basic functions such as multi-level catalogs, Markdown support, and offline synchronization optimization that users urgently needed was put on hold, and flashy functions such as “handwriting recognition” and “intelligent scanning” were piled up.
What’s worse is the management turmoil, the CEO changes frequently, and the strategy swings back and forth. Evernote’s CEOs are either obsessed with the vision of “humanity’s second brain” and neglect commercialization; or radically turn to the enterprise-level market, but lack product support.
In 2017, Evernote’s board of directors rejected a $3 billion offer to sell to Microsoft, missing the best time to exit.
How can product managers do a good job in B-end digitalization?
All walks of life have taken advantage of the ride-hail of digital transformation and achieved the rapid development of the industry. Since B-end products are products that provide services for enterprises, how should enterprises ride the digital ride?
View details >
Later, coupled with the market being segmented by competitors such as Notion, Slack, Obsidian, and Craft Docs, Evernote’s market position gradually became marginalized, and its valuation continued to decline.
In 2023, Evernote was acquired by Italian mobile app developer Bending Spoons, and all employees were laid off, and the product was completely taken over by Bending Spoons employees.
A small but beautiful legend came to an end.
01 The fate of tools
Either being cannibalized by emerging competitors or being acquired is a fate that small but beautiful tool-based software is difficult to get rid of.
Especially since the wave of large models, the reshuffle of tool-based software has been accelerating. Niu Reuters has compiled a list of tool-based software affected by the reshuffle in recent years (partial), in this reshuffle, there are many unicorns with a valuation of tens of billions of dollars:
As early as the mobile Internet era, there were many tool-based software that was acquired.
From the logic of acquisition, large manufacturers are either to fill the ecological gap, or to penetrate their business into vertical industry scenarios, or to absorb each other
Core technical team.
The outcome of the acquisition of tool-based software is nothing more than the disappearance of software brands from the market like Teambition and Meiqia; or live as a sub-module of Tencent Cloud like Coding and become a vassal of a large company.
Why did this result occur?
Niu Reuters combed through the development history of these brands and found that in addition to common problems such as low user stickiness, low payment conversion rate, and strategic mistakes, there is also an almost common problem – thin functions and easy to replace. This is also an important reason why many tool-based software are eventually acquired or eliminated.
The advent of generative AI has revolutionized the workflows of businesses and organizations.
Traditional software tools are usually oriented to specific links, such as Excel processing data, PS retouching, etc., and require manual connection of multiple tools to complete tasks; In the era of AI large models, an agent can autonomously complete the entire process from task understanding, execution to result output.
For example, when generating a market analysis report, AI can automatically complete data collection, analysis, visualization, and copywriting, eliminating the need for manual step-by-step operations.
The role of people has changed from executives to supervisors, workflows have changed from human-driven to AI-driven, and tool-based software vendors have changed from providing tools to directly delivering results, which is the most essential change in tool-based software in the AI era.
After Internet giants and leading manufacturers of large AI models launch AI applications, those small and beautiful software with single functions may be hit by dimensionality reduction.
Under the changing situation, how should tool-based software manufacturers survive?
02 All in One, or vertical deep plowing?
Many manufacturers choose “All in One” to build a stronger product matrix through integration to resist external competition.
Dr. Fan Ling, founder and CEO of Tezan Technology, mentioned in an interview with Niu Reuters that Tezan has gradually expanded from digital asset management to full-link services for content production, distribution, and analysis, and this extension of the tool chain is not accidental, but a backward pressure from customer needs – enterprises no longer need a single tool, but a solution that can run through the entire life cycle of content.
By building a closed-loop ecosystem of “content + AI”, Tezan seamlessly connects all aspects of the content life cycle in order to gain an advantage in the competition.
(Picture) The “content + artificial intelligence” system created by the brand is specially praised
Similarly, AI office incubator brand PixelBloom has adopted a similar strategy.
Pixel Bloom first started with the graphic layout tool “365 Editor”, and later continued to enrich its product line, successively launching the graphic design editor “Love Design”, the digital asset management platform “AIGC Content Center”, AiPPT.com, AiH5, Visdoc.cn, Aibiao.cn and other products.
PixelBloom (AiPPT.com) Founder & CEO Zhao Chong told Niu Reuters that the future office software competition is no longer about a single function, but about ecological collaboration capabilities.
When a product reaches the ceiling of commercialization, PixelBloom will develop new products and expand the imagination of the business.
(Figure) PixelBloom’s development history and productivity tool product matrix
Both companies are integrating toolchains to reduce the cost of switching between different platforms while enhancing the irreplaceability of their products.
Of course, there are also manufacturers who are not moved by the trend and insist on deeply cultivating a single product.
After the outline note-taking tool brand “Curtain” was acquired by Byte, it sold Flomo ink notes again, but Flomo refused to “All in One”, did not integrate or connect the screen with the original product, and maintained the independent operation of the screen.
Flomo is a tool that focuses on lightweight note-taking, and they don’t do all-in-one, but insist on “recording and organizing ideas”, which has also won a loyal following.
The founder Liu Shaonan’s philosophy is that the value of a tool does not lie in the number of functions, but in whether it can solve a certain problem to the extreme, less is more.
Perhaps Liu Shaonan saw the potential risks of the “All in One” strategy, where overexpansion could lead to bloated product functionality and weaken the core experience.
After all, Evernote has learned a lesson from trying to maintain its innovative image by expanding edge features such as food records and e-commerce, only to be abandoned by users because they neglect core needs.
Does “All in One” really meet the needs of users? What kind of tools do users need?
Xiao Yijun, CMO of Jingshuo Technology, shared their use of AI tools with Niu Reuters. Their experience is that AI tools perform well in scenarios such as short copy generation and content reuse, but they still need manual intervention in brand tone and in-depth content creation.
Xiao Yijun’s team prefers to use agents that connect tools, that is, AI Agents that are tuned according to tasks in different scenarios. For example, SEO Agent, SDR Agent, etc. This allows you to connect to your own knowledge base, and the output availability is much higher than that of the general version.
They also use different tools in series, such as video production, combining clipping, dreams, and in-house AI agents instead of relying on an all-in-one product.
Zhang Fan, brand director of Panneng.com, also attaches great importance to the professionalism and brand safety of tools. In an interview with Niu Reuters, she said that the in-depth functions of independent tools in some unique industry verticals are still difficult to be replaced by AI applications built into large models, especially in copyright-sensitive enterprise scenarios, where users are more inclined to choose mature and compliant professional tools.
It can be seen that users do not blindly pursue “All in One”, but value the professionalism and ease of use of tools in specific scenarios.
Of course, tool manufacturers must not only follow the logic of users, users only choose products that are easy to use, manufacturers must also consider competition in the same industry, and consider building a moat.
03 Make yourself “irreplaceable”
Leading manufacturers are accelerating the embedding of AI capabilities into various applications with their technology, data, and ecological advantages.
Microsoft Copilot is integrated into Office, DingTalk is connected to Tongyi Qianwen, and WPS is launched to generate AI-generated PPTs…… These “buckets for the whole family” solutions are squeezing the living space of standalone tool-based software.
In the face of the dimensionality reduction blow of large manufacturers, how should small and beautiful tool manufacturers break through?
A founder of enterprise service software told Niu Reuters that AI solutions from large manufacturers often pursue “big and comprehensive” in an attempt to meet the basic needs of the widest range of users.
This universal design creates a living space for tool manufacturers who focus on subdivided fields – through the vertical cultivation of “specialization, refinement, and innovation”, small and beautiful tools can establish differentiated advantages that are difficult to replace in specific fields.
Dr. Fan Ling led the team to develop atypica. AI business research agents are an example of vertical deep cultivation.
Unlike most AI tools on the market that directly call general-purpose large models, atypica. AI is based on Tezan’s self-developed “Creative Reasoning” reasoning framework, designed to solve a wide variety of business problems.
Business problems are usually “Wicked Problems,” which are open-ended questions that do not have standard answers and require complex reasoning, such as predicting market trends, analyzing consumer behavior, etc.
“General large models are good at solving problems with clear answers, such as writing a business email, but when enterprises need to decide whether to enter the Southeast Asian market, they need to combine industry data and consumer insights for multi-dimensional deduction.” Fan Ling said.
Fan Ling believes that large language models solve language problems, and when language problems are solved, thinking problems can be solved, and thinking problems can be solved, and what kind of thinking process a person will have on a certain problem.
atypica. AI uses it to build consumer personas by browsing through various social media data in real time. After establishing consumer personas, interviews are conducted with these personas through the surveyed agents, and then the interview minutes are integrated to deduce some possible answers to the research questions.
By simulating human unstructured thinking, combining qualitative analysis with quantitative data to generate practical business proposals, this deep vertical ability is difficult to replicate in the short term with general tools.
Zhao Chong said frankly that their core competitiveness does not lie in the basic function of “being able to generate PPT”, but in “better understanding the PPT needs of Chinese users”.
In the Chinese market, the style requirements of PPT vary greatly among enterprises in different industries and natures: state-owned enterprises need solemn and atmospheric design, Internet companies prefer minimalist styles, and consulting companies emphasize data visualization.
AiPPT.com analyzes 200,000 industry templates and user behavior data to automatically adapt to these subtle but critical stylistic differences.
“Our AI not only knows how to typeset, but also knows which colors to avoid in central enterprise reports and how to highlight core data in roadshow PPTs.” Zhao Chong said.
This deep understanding of subdivision scenarios comes from the team’s long-term accumulated industry know-how and continuously optimized exclusive models.
To establish such a vertical advantage, tool vendors must focus on a sufficiently vertical and demanded scenario, with continuous data accumulation and model optimization, and must provide deep capabilities that are difficult for large manufacturers to replicate.
04 Complement giants, not compete
Instead of fighting against the ecology of large factories, it is better to actively integrate and become an indispensable part of its value chain.
Embedded ecological cooperation is also an art of survival.
AiPPT.com will soon be deeply embedded in DingTalk documents via API. When users edit documents on DingTalk, they can directly call AiPPT.com’s engine to generate professional slides, and users don’t even realize that they are using third-party services.
AiPPT.com is also pre-installed in 50 million commercial computers around the world, including Lenovo and HP, and can be used by users as soon as they turn it on. This also brings a stable traffic entrance to the tool, and this ecological cooperation has contributed many new users to PixelBloom.
“Large model manufacturers themselves have built-in PPT generation functions, why do users choose us? In fact, frankly speaking, more than 70% of the built-in PPT functions you see are our ability to access them. Zhao Chong revealed to Niu Reuters.
He said that if you want to win the favor of users, there are two core points: professionalism and openness. They have been working in the field of PPT for three years, accumulating more than 200,000 sets of template resources for different industries, different occupations, different scenarios, and even cultural adaptation in different countries.
Fan Ling used cars as a metaphor for the current AI ecosystem: “Large models are the engines of the new era, but a car cannot be solved by just one engine.” He suggested that entrepreneurs should focus on the seats of the car, the experience of the car, and the driving style…… and so on.
In his opinion, investing 80% of resources in scenario-based innovation and experience optimization is a wise choice for startups.
05 What is a moat?
What is the moat of tool-based software?
Many people think that the moat of technology lies in patents or algorithms, but for tool-based software, the real barrier is “vertical data + engineering capabilities + ecological network”.
The first is data. Zhao Chong took PPT generation as an example: “We process a large number of user requests every day, and this data tells us the difference in demand in different industries – the education industry needs a clear courseware structure, and the financial industry relies on accurate data charts. This feedback continuously optimizes the model to make the tool more user-wise. ”
This kind of scenario-based data is difficult to be quickly replicated by large manufacturers and has become a core barrier in subdivided fields.
The second is product engineering capabilities. Technology is only the foundation, and if you want to win the market, the most important thing is speed, how to quickly discover user needs, quickly discover market opportunities, quickly productize, quickly PMF test, quickly lay channels, and quickly make profits.
Finally, there is the ecological network. AiPPT.com cooperates with hardware manufacturers, software platforms, and content copyright owners to form a closed loop, and once users enter this ecosystem, it is difficult to leave.
For example, if you use their tools to generate PPT on Lenovo computers, the material comes from Visual China’s 500 million copyright gallery, and then share it directly with collaborative colleagues.
Therefore, the moat is not a single technology, but a superposition effect of these elements.
Fan Ling’s proposition is not to fight for technology with large factories, but to fight for efficiency and creativity.
“The advantage of a startup is never pure technology, but efficiency and flexible iteration”, Fan Ling pointed out, “The road is out, the end point of a startup is often different from the starting point, the key is to adjust quickly, not a BP (business plan) to the end.” ”
06 Epilogue
The dilemma of tool-based software is essentially the collision of single-function thinking in the industrial era and the systematic and intelligent needs of the AI era.
The decline of Evernote, the transformation of Tezai, and the ecological breakthrough of Pixel Bloom all show that the value of tools is shifting from “problem-solving” to “defining problems” in today’s AI restructuring workflows.
The moat is no longer a technical patent or user scale, but a closed loop of data in vertical scenarios, the speed of engineering implementation, and the integration with the big ecosystem.
The window of time left for independent tools is closing, and these products will either become irreplaceable “specialized, specialized, and new” links on the ecological chain, or become rigid in the past vision like Evernote.
